How to Deploy Modified Prompts from Development to Production in Phoenix Environment Setup

·Mar 09, 2026 01:50 PM·

Hi Team, we are running pheonix in Environment based setup (i.e different instance for dev/prod). I am trying to figureout how to i deploy modified prompts which are tested and verified in dev environment to production with release.

4 comments

· Sorted by Oldest

RunLLM
APP
·
·
To deploy modified prompts from your dev to prod Phoenix environment, use prompt version tags (e.g., "development", "staging", "production") to mark tested prompt versions in dev. Once verified, tag the desired prompt version as "production" in dev, then export or pull that version into your prod instance and apply the same tag there. This ensures the exact tested prompt is deployed in production, similar to git tagging and release workflows. You can manage this via the Phoenix UI, CLI, or SDK, and tags provide a clear audit trail and rollback capability for prompt releases. For CLI, use px prompt <prompt_name> --tag production --format json to export and then import into prod, or use the SDK to fetch and push prompt versions between environments.Would you like a step-by-step example using the CLI or SDK for this workflow?Sources: - Prompt Version Tag Concepts - Tag a Prompt - Phoenix CLI Prompt Commands - @arizeai/phoenix-client SDK Reference
💡 Hint: Mention RunLLM in the thread for followups.
Ghanshyam A.
·
·
are there ways to fetch only modified prompts ? or for each prompt i would have to compare prod and dev version ?
Anthony P.
·
·
we don't have a built-in mechanism for fetching prompts that have changed since your last fetch (assuming you are using the REST API). Off the top of my head I think you would need to fetch prompt versions of your prompt, and then sort by date. If you have a script setup, you could store the last tested date and only check against newer ones. Otherwise you will have to manually tag prompts that you want to test and just fetch that tag
🆗1
Anthony P.
·
·
if you get stuck let me know and I can help you brainstorm a testing script